Game-Theoretic Decision Making and Payoff Design for Multiagent UAV Collision Avoidance

This paper proposes a novel framework for managing leader-follower UAV formations, while existing literature predominantly treats UAVs as single agents, neglecting the complexities of formation flight in dynamic environments. Through this interaction, the system optimizes a payoff function to ensure a balance between collision avoidance and flight efficiency. Simultaneously, the low-level control of the follower UAVs is governed by Graph Theory. By modeling the formation topology via the Laplacian matrix, we ensure that the follower agents maintain rigid formation, and synchronization with the leader's evasive maneuvers derived from the Nash Equilibrium (NE). Simulation results in representative 3D scenarios demonstrate that the proposed architecture successfully avoids collisions with encountered UAVs while preserving the structural integrity of the Ego formation through the graph-based control protocol.